Hallo,

I also took a look at your model and I have some suggestions in addition to what Jules wrote:

1. Use symmetry condition, i observed that you are not simulating for the middle domain (which i find weird). However, think about using symmetry to decrease the size of your mesh or to use a finer mesh with same computational resources.

2. I have built the mesh using physics-controlled mesh, and the mesh quality looks ok. If you suspect that the mesh is the problem then you can try and run a mesh refinement study on the model. Just google: comsol mesh refinement study.

3. I observed the log of your previous solution and noted that you first didn't define a range for your outlet, only range(0,23200)s. As such, and due to other reasons you see in the log a lot of "-out" between different study step. This -out means that the solution is interpolated instead of numerically solved. This can be the cause for the unsatisfying result you obtained. In which, the changes in the system are not correctly tracked in the solver. One way to bypass this problem is by going to: time-dependent solver 1>settings>time stepping>steps taken by solve: Strict or by changing the maximum step to e.g. 2 seconds.

I hope that helps,

Regards,

Zaid

Hello Mohammed,

When I open your model I do not see any saved mesh to investigate? More importantly I cannot seem to run the model as I have COMSOL reporting an error with the 'aveop1' operator.  COMSOL is also reporting several warnings of unknown variables- such as global variable k_fun and your i_v reaction coefficients for the porous electrode couplings.

Whilst you have stipulated an inlet velocity (v_dot/(ch_H*ch_W)), you have not set the boundary! The same goes for the outlet! 

I would recommend either carefully going through your model and checking for basic errors or starting with a much simpler model, such as a simple 2D representation of your problem.

This may be of use: http://www.comsol.com/community/forums/general/thread/42221/

Also have you calculated the Reynolds number for your system to determine the flow regime?

I have also provided some links to understanding and improving your mesh that I hope will be useful to you to solve your initial question on meshing.

Hopefully someone here can help more.

Kindest regards,

Jules

http://www.comsol.com/blogs/meshing-your-geometry-various-element-types/

http://www.comsol.com/blogs/improving-your-meshing-with-swept-meshes/

http://www.comsol.com/blogs/improving-your-meshing-with-partitioning/

https://www.comsol.com/blogs/using-adaptive-meshing-local-solution-improvement/

Hallo,

I also took a look at your model and I have some suggestions in addition to what Jules wrote:

1. Use symmetry condition, i observed that you are not simulating for the middle domain (which i find weird). However, think about using symmetry to decrease the size of your mesh or to use a finer mesh with same computational resources.

2. I have built the mesh using physics-controlled mesh, and the mesh quality looks ok. If you suspect that the mesh is the problem then you can try and run a mesh refinement study on the model. Just google: comsol mesh refinement study.

3. I observed the log of your previous solution and noted that you first didn't define a range for your outlet, only range(0,23200)s. As such, and due to other reasons you see in the log a lot of "-out" between different study step. This -out means that the solution is interpolated instead of numerically solved. This can be the cause for the unsatisfying result you obtained. In which, the changes in the system are not correctly tracked in the solver. One way to bypass this problem is by going to: time-dependent solver 1>settings>time stepping>steps taken by solve: Strict or by changing the maximum step to e.g. 2 seconds.

I hope that helps,

Regards,

Zaid

Hello,

Like mentioned in the previous two posts, there are some mistakes in your model.

In the parameters section, "k_fun" has an invalid unit (you have set the unit of k_fun to be [m/s] while you're equation suggests it should have units of [velocity^(0.4)]. If you want k_fun to be dimensionless modify your equations to be "1.6e-4*(v[s/m])^0.4"). While this shouldn't cause any significant errors in the simulations, no harm in fixing it.

I too have meshed your model with a physics controlled mesh and it looks fine on the first look. Like mentioned in the previous post you have lot of "-out" in your previous solution. I don't think this is due to an issue with your mesh. If you believe it's a problem with the mesh, you can try different types and refinements of mesh and see if the solution improves. Try changing time stepping to "strict" and see if it improves your solution.

I cannot solve your model because you have lot of unknown variables in your global equations.

Good luck

Regards,

Saliya.

Jules, Zaid, and Saliya   thank you so much for your valuable comments and actually you added new things to me.

First of all, sorry for making you confused because I did no post the entire model. Secondly, I tried with what you have mentioned by using , physics-controlled mesh, but it shows me a ‘’ COMSOL Error ‘’ as attached in pic. 1, and when I hit the button ‘’ OK ‘’ it shows me the meshed geometry as attached in pic. 2. As you see in pic.2, beside the mesh being very fine, there is an error under ‘’ problem 1, at the L.H.S. and it did show me any solution after running it because of this error.

So what I did is deleting the old mesh and adding new one, then by using ‘’ Free Tetrahedral 1’’ I hit the ‘’Build All’’ button and I got the mesh but I think it is very coarse and not uniform as shown in pic. 3. Eventually, I run the problem and I got results as shown in pic. 4, but as you see the colors on the geometry, specifically the top surface of the channels, are not uniform even if at time zero where the initial condition is applied and the colors and values should be same in this case and the reason as I think is the mesh where it is very coarse in the channel domains, specifically at the inlet and outlet surfaces.

In fact, what I want is to mesh the whole geometry with reasonable meshing, to avoid any kind of these mismatching colors after getting the solution like what I mentioned above.

Also, Jules, I calculated the ‘’Re’’ number and it was 133 and 70 for both channels and this means the regime is laminar, but what if it is turbulent?

By the way, is the way of building the geometry correct? Also, after checking the ‘’ log’’ I noticed a lot of ‘’ out’’ NOT ‘’ ‘’- out’’, is that fine, please?

Thank you again and I hope I can get another chance for helping with this meshing problem.

Best regards,

Mohammed

Hello,

If you want to improve the mesh and see if it changes your result, you can goto the size node under the mesh and change the size from normal to fine, extra fine, etc. Additionally, when you add a free tetrahedral mesh to your model, by default it's calibrated for general physics. Under the size tab you can see this. By changing this to Fluid dynamics and setting the size to normal, you can get a better mesh for your model.

If you want to refine your mesh further, you can define different mesh sizes for different elements(domains, boundaries...) in your geometry. You can do this by right clicking on Free Tetrahedral and adding extra size nodes. Once you add additional size nodes you can choose different  domains, boundaries etc for each of those size nodes and set the mesh sizes individually for those selected geometry levels.

Regards,

Saliya

Hallo,

First of all, update the version of comsol if possible. I see that you are using v.4.3. The thing is, i am using comsol v.5.1 and i strongly believe it is more stable than its predecessors. The meshing algorithm is more stable than the version you are currently using and therefore, the other colleagues and i dont get any meshing errors.

In case you are not able to upgrade the version, then you have to try manual meshing, which is both time-consuming and rather frustrating (telling you from experience). You require the boundary layers to solve for the flow near to the wall boundaries, so meshing using only free tetrahedral is not sufficient.

So in case you want to manually mesh it, i suggest you first start with the channels:

1. Free tetrahedral>domain selection: channel domains>right click and size> normal (fluid dynamics).

2. Add boundary layers node> choose same channel domain>right click boundary layer properties> choose exterior wall domains of the channel (except inlet and outlet) and add 3 layers.

3. Add Free tetrahedral> remaining geometry.

With some luck, i am sure that the mesh would now work.

Another thing concerning the units. It is not of importance, the highlighted text only states that the unit must be a predefined SI value. For example: for velocity is m/s. I have this case sometimes where the unit is not similar to the required one. If you are sure of your equation, then dont pay it any attention since it is not a source of error.

To answer your question, yes, your regime is laminar and i dont suspect you will have any turbulence due to the channel geometry. So the mesh i described above would be sufficient, otherwise extra effort must be made to define the boundary mesh layers.

Another thing: -out is the same as out... sorry for the confusion, it might be comsol version related since i only see -out.

Finally, your results, i dont know what you expect as a result, but what i observe from your attached picture is that:

1. Concentration value lies around 156 mol/m^3, therefore the maximum and the minimum value are both showing a constant of 156.

2. there is something happening in the simulated model, but the mesh is indeed coarse, you can observe straight lines around the result area (not a smooth result). However, it is not of major concern.

3. The time for the surface concentration plot is not shown, so i am not sure if that is at the end or the beginning of the simulation.

So to evaluate your result, try to change the range of the data from the surface plot> manual color range.

Finally, try to play around with your simulation parameters, see if you get other results. Maybe you will find out something new about your model.

Good luck and best regards,

Zaid

Thank you again guys for spending time and helping me with this problem and I will try to follow your instructions.

Zaid, if I am going to update my version, does the default meshing sufficient or not?

The posted results at time = zero where the initial concentration is 156, so it makes sense.

By the way let me tell you something, I solved exactly the same problem, vanadium redox flow battery model, by using the same parameters and physics, BUT without adding channels, only three blocks, and I got very reasonable results. However, adding the channels still showing me results but the weird thing as follows.

 For example, the required time to reach 1.7 Volt during the cell charging is 1700 seconds with adding channels while 13000 seconds without adding channels( this one  makes sense). So, this is why I guess the issue with adding the channels is coming from the unsuitable meshing.

The only difference between the case with and without adding channels is the velocity value because at same flow rate and at different cross section areas the velocity will be different, 0.004368 m/s without channel and  0.135 m/s with channels, but the regime still laminar.

Hopefully this will give you more idea about my actual issue after adding the channels to the 3 blocks.

Thank you and best regards,

Mohammed

Hallo again,

I believe that the default mesh is sufficient in the new comsol version. Mesh>right click: Statistics, the histogram must look similar to a normal distribution oriented to the right side. This is an indication of a good mesh.

If the solution is at time = 0, then the solution is ok. Observe always the maximum and minimum value of the color bar. The reason behind color plot shown in time = 0s is the fluxes at the boundaries which causes minute changes in the concentration.

I am no specialist in your field. However, i can tell you that a simulation output change observed after 13000 s (without channels) and 1700 (with channels) is not mesh related. The difference it almost x10, which cannot all be due to the mesh. The nexxt best thing to do is to validate the model.

I wish you luck in finding the reason for this deviation.

Regards,

Zaid

 Thank you so much all.